Practical Approaches to Uncertainty Analysis for Light Measurement Systems Using Integrating Spheres and Spectroradiometers

Transcription

1 Practical Approaches to Uncertainty Analysis for Light Measurement Systems Using Integrating Spheres and Spectroradiometers Dan Scharpf Labsphere, Inc. CORM 2015 Annual Conference Boulder, CO May 12-15, 2015

2 Outline Background Uncertainty Overview The Integrating Sphere / Spectroradiometer System Uncertainty Contributions Summary

3 Uncertainty Overview I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it. Why? - Lord Kelvin Because we have to. IEC/ISO17025 NVLAP HB-150 Because we need to. Sets a level playing field. Identifies areas of concern Can identify changes in process and gauge quality

4 Uncertainty Overview Case Study References: J. Leland, Uncertainty Analysis for an Integrating Sphere Photometer - A Case Study, CORM 2014 A. Jackson, Case Study: Calibration Uncertainty Analysis for an Integrating Sphere Spectroradiometer, CORM 2014 R. Bergman, M. Paget, E. Richman, CALiPER Exploration Study: Accounting for Uncertainty in Lumen Measurements, USDOE, PNNL March 2011 Uncertainty Methods: D. Gross, Interpolation, Correlation and NIST Standards, CORM 2014

5 Uncertainty Overview 1) Define the fundamental equation: (measurand) 2) Calculate the Sensitivities: C xi = f x i 3) Determine the Uncertainties: u x i y = f x 1, x 2, x 3, x n f x i x 3 x 2 x 1 4) Calculate the Coverage Factor: k 0.95 x i Law of Propagation of Uncertainties: Combined Standard Uncertainty u c 2 y = Sensitivity Contribution; add in quadrature (RSS) n i=1 f x i 2 u 2 x i + 2 n 1 i=1 n j=1+1 Correlation Term f f u x x i x i, x j j

6 Uncertainty Overview How do we get the uncertainty estimates for each term? Measure it acquire statistical data Estimate it based on literature Calculate it numerically Analytical/Empirical model Product data sheets

7 Integrating Sphere / Spectroradiometer: System 1 Auxiliary Lamp Calibration Lamp DUT Power Supply Optical Fiber Spectroradiometer DC Constant Current DC Constant Current Calibration Lamp Power Supply Auxiliary Lamp Power Supply

8 Integrating Sphere / Spectroradiometer: System 2 AC with DC Shunt Resistor AC Power Analyzer DUT AC Power Supply Optical Fiber Spectroradiometer Aux Lamp DC Power Supply Precision Shunt Multimeter Calibration Lamp DC Power Supply

9 Calibration Procedure: 4 Measurements Configuration Measurement 1. Reference Standard On y R (l) 2. Aux. Lamp On, Ref Lamp Off y xr (l) 3. Aux Lamp On, Empty Sphere y xe (l) 4. Aux Lamp On, DUT in place y xdut (l) 5. DUT On (Aux off) y DUT (l)

10 Measurement Details A dark scan is taken prior to each measurement to remove the baseline offset from the scan. e.g., for the Ref. Lamp: y R λ y Rd λ All Scans are wavelength dependent. Spectral range depends on Ref. Standard and Spectrometer Spectrometer Manipulations: Pixel Interpolation to get integer wavelength values Stray Light Correction Counts conversion to engineering units Applying Calibration Certificate Values Interpolation

11 Uncertainty Overview: Measuring Spectral Flux φ DUT λ = y DUT λ y DUTd λ y R λ y Rd λ φ R,λ λ L e λ, T L e λ, T 0 y xr λ y xrd λ y xdut λ y xdutd λ Scale Factor Ref. Lamp Calibration Spectral Shift Correction Aux Correction L e λ, T 0 = Spectral Radiance at T 0 L e λ, T = Spectral Radiance at T R= Ref Scan d = Dark Scan x = Aux Scan DUT = Device Scan L e (l) = Spectral Shift

12 Uncertainty Overview: The Sphere Calibration Coefficient φ DUT λ = y DUT λ y DUTd λ y R λ y Rd λ φ R,λ λ L e λ, T L e λ, T 0 y xr λ y xrd λ y xdut λ y xdutd λ Isolate the terms from the calibration of the system, normalize to an empty sphere, include correction terms. C φ λ = y R λ y Rd λ φ R,λ λ L e λ, T 0 L e λ, T y xe λ y xed λ y xr λ y xrd λ A R C fr A R = Aging Factor C fr = Additional Contributions

13 Uncertainty Overview: Using the Sphere Calibration Coefficient φ DUT λ = y DUT λ y DUTd λ C φ λ y xe λ y xed λ y DUTAUX λ y DUTAUXd λ Measurement of DUT utilizes an Aux-Empty Scan.

14 Sources of Uncertainty Noise (Type A) C φ λ = y R λ y Rd λ φ R,λ λ L e λ, T 0 L e λ, T y xe λ y xed λ y xr λ y xrd λ A R C fr Look at each term separately and identify sources of uncertainty for each term.

15 Sources of Uncertainty Variable Impact Uncertainty Source φ R,λ λ Spectral Flux L e λ, T 0 L e λ, T Flux Spectral Shift Total Flux Spectral Resolution Ref. Lamp Uncertainty Lamp Current Ambient Temperature A R Aging Factor C fr Additional Terms Flux Flux Lamp Decay Spectrometer Stray Light Light Leakage into Sphere Near Field Absorption Sphere Non-uniformity Spectrometer Nonlinearity Ambient Temperature

16 Uncertainty Contributions: Measured Values C φ λ = y R λ y Rd λ φ R,λ λ L e λ, T 0 L e λ, T y xe λ y xed λ y xr λ y xrd λ A R C fr 1. Take N scans (N=3 to 10) 2. Calculate the mean 3. Calculate std. dev. 4. u y λ y d λ = Std.Dev. Mean Measuring multiple scans of the reference lamp includes: Spectrometer noise Power Supply noise Dark Subtraction

17 Uncertainty Contributions: Ref. Lamp Calibration Certificate, φ R,λ λ C φ λ = y R λ y Rd λ φ R,λ λ L e λ, T 0 L e λ, T y xe λ y xed λ y xr λ y xrd λ A R C fr 1. Interpolation is required to get 1 nm interval data from 10nm report. 2. Spectrometer wavelength accuracy influences certificate values 3. Transfer uncertainty is reported on calibration certificate.

18 Uncertainty Contributions: Ref. Lamp Calibration Certificate, φ R,λ λ C φ λ = y R λ y Rd λ φ R,λ λ L e λ, T 0 L e λ, T y xe λ y xed λ y xr λ y xrd λ A R C fr Uncertainty from Interpolation and Wavelength Accuracy 1. Interpolate φ R,λ λ and u φref data from cal. Cert. 2. Calculate φ R,λ λ + Δλ 3. Calculate the difference: Δφ R,λ = φ R,λ λ + Δλ φ R,λ λ 4. Calculate the Relative Uncertainty: u φwavelength = Δφ R,λ φ R,λ Combined Uncertainty: RSS the values: u φ = 2 2 u φref + u φwavelength NOTE: See A. Jackson, Corm 2014, for more rigorous approach to include uncertainty between data points.

19 Relative Uncertainty, % Spectral Flux, W/nm Uncertainty Contributions: Ref. Lamp Calibration Certificate, φ R,λ λ Reference Lamp Spectral Flux Cubic Spline Interpolation 0.06 Spline Data, W/nm Shifted Data, W/nm 0.04 Cal Cert Wavelength, nm Reference Lamp Uncertainty Total Relative Uncertainty, % Calibration Certificate Uncertainty, % Wavelength Accuracy Uncertainty, % Wavelength, nm

20 Uncertainty Contributions: Spectral Shift Exponential Fit Method C φ λ = y R λ y Rd λ φ R,λ λ L e λ, T 0 L e λ, T y xe λ y xed λ y xr λ y xrd λ A R C fr Spectral shift for a tungsten halogen lamp has been related to the blackbody radiation curve as: L e λ, T 0 L e λ, T n J JR exp a 1 = J ref exp a 1 a = hc λ kt 0 h = Planck s Constant 6/ E-34 m 2 kg/s c = Speed of Light, E+13 nm/s k = Boltzman Constant, E-23 m 2 kg/s 2 -K T 0 = Color Temp from Ref. Lamp Calibration Cert., K J R = Reference Current from Calibration Cert., A J = Current during Sphere Calibration, A n JR = CCT Sensitivity Exponent, 0.648

21 Uncertainty Contributions: Spectral Shift Exponential Fit Method L e λ, T 0 L e λ, T n J JR exp a = J 1 ref exp a 1 a = hc λ kt 0 Variable Source Uncertainty Source J = 2.680A n jr = Current During Sphere Calibration Calculated by measuring flux at different currents. Ref. 1, A Spec Sheet Calculated Ref. 1, C. Miller, NIST measured data Ref. 2, A. Jackson, CORM methodology

22 Uncertainty Contribution: Lamp Power Supply Current Check the Spec sheets: Labsphere LPS-150 Current accuracy = ±0.1% Temperature Coefficient: ±0.1% / C at 25C u i = 0.001i lamp (Tamb 25C i lamp 2 = A Agilent E3634A Depends on manual/computer setting Depends on range used Current Accuracy Good for 25C ± 5 C Programming: ±(0.2% + 10 ma) Readback: ±(0.15% + 4mA) u i = i lamp A = A For 22C

23 Uncertainty Contributions: Spectral Shift Blackbody Method From theory and measurements, a 1% change in lamp current produces a 0.8% change in blackbody temperature and a 2.9% change in total power. Use Planck Blackbody function to calculate spectral flux at T. Total flux is given by Stefan-Boltzmann Equation. Planck s equation : φ λ T = 2πhc2 λ 5 1 e hc λkt 1 W μm Stefan-Boltzmann: φ T = σt 4 W lamp Spectral Flux per 1 W of Lamp Power at T: F λ T = 2πhc2 λ 5 1 e hc λkt 1 1 σt 4 W μm W lamp

24 Uncertainty Contribution: Spectral Shift due to Lamp Current Calculate the Sensitivity of Spectral Flux due to spectral shift from lamp current uncertainty: 1. Calculate F l (T) at CCT of Cal lamp, T cal. 2. Calculate F l (T) at x T cal. 3. Change in Spectral Flux for a 1% change in current due to spectral shift: 4. ΔΦ λ = F λ T F λ T T 5. Calculate the change in flux due to the uncertainty in the current: 6. u φ λ = u i i cal 100 ΔΦ λ (C) 2015 Labsphere, Inc. Proprietary and Confidential

25 Uncertainty Contribution: Spectral Flux due to Cal Lamp Current The change in total power is 2.9% for a 1% change in current: 1. Calculate additional power due to uncertainty in current: 2. u φ λ = u i i cal φ 3. Total change in flux is the sum of each contribution. 4. u φ λ = u i i cal 100 ΔΦ λ + u i i cal φ

26 Power, Watts Uncertainty Contribution: Total Flux due to Cal Lamp Current Estimate the change in flux due to lamp current by measurement. 1. Step lamp current between Nominal ± 3% 2. Measure Spectral Flux at Each Current Setting 3. Plot variation in total flux 4. φ i = W/Amp 5. Uncertainty in flux due to lamp current: u φi = φ u y = x i i 20.0 R² = Lamp Current, Amps

27 Uncertainty Contribution: Spectral Flux due to Cal Lamp Current Estimate the change in flux due to lamp current by measurement. 1. Step lamp currents for Nominal ± 3% 2. Measure Spectral Flux at Each Current Setting 3. Plot variation 4. Calculate φ i 5. Uncertainty in flux due to lamp current: u φi λ = at each wavelength φ λ i u i

28 Uncertainty Contribution: Spectral Flux due to Cal Lamp Current % change in Spectral Flux for 1% change in Lamp Current.

29 Uncertainty Contribution: Lamp Aging Measure it: Model it: At 120% of lamp life, 8% degradation in flux. u Lifetime % = Hrs Used Lifetime 120% 8% 13 Ref: Labsphere Lamp Depreciation Tech Note.

30 Stray Light into Sphere Measure it 10 scans with no system lamps on. Scans are relative to the signal from the cal lamp Normalize with integration time Divide by 3 for rectangular distribution u stray light λ = y dark y ref t dark t ref 1 3

31 Near Field Absorption At close distances the surface may not produce a Lambertian reflectance. The amount of light absorbed by surfaces that do not have reflectance equivalent to the overall sphere. Estimate the amount of light that strikes the holder. Assume 5% of the light interacts with near-field, f nf Assume 10% of this light is absorbed, α nf, reducing the measurement: u nf λ = f nf α nf 100% = 0.5%

32 Sphere Non-uniformity Measure it: Rotate cal lamp in 5 increments about center post. Measure Radiant Flux and normalize by mean value. From previous data, we measured Coeff. of Variation in Radiant Flux for 10 stationary scans = 0.09%. Variations > 0.9% influence of uniformity. Use a Constant value: u nu = 0.2% Function of Wavelength: Std. Dev at each wavelength across the rotation angles. u nu λ 1.48E 4λ E 2, 350 < λ < 400 nm u nu λ 0.2%, 400 < λ < 1050 nm

33 Spectrometer Stray Light Errors due to light reaching unintended pixels in spectrometer. Greatest impact at low wavelengths when measuring blue LEDs after calibration with QTH lamp. Measurements indicate 1% impact in radiant flux at for measuring blue LEDs from QTH calibration. Minimal effects for l > 450nm for all lamps. u sl λ u sl λ = 1%, nm for Blue LED = 0 for l > 450 nm.

34 Spectrometer Nonlinearity Effect of measuring a device whose output is orders of magnitude different than the calibration lamp. Define the photopic ratio: P. R. = Photopic ratio = φ DUT φ Cal lamp Base the uncertainty on the dynamic range of the spectrometer: u nl = 0.1 u nl = 0.3 log P.R. log 60 log P.R. log 60 for CDS-1100/2100 for CDS-600/610

35 Summary: Inputs required Ref. Lamp Calibration Certificate Color Temperature Lamp Current Radiant Flux Component information Spectrometer Spec Sheet Dynamic Range Wavelength Accuracy Stray Light information Reference Lamp Hours Used Power Supply Spec Sheets Accuracy Temperature Dependence Sphere Uniformity Near Field Absorption Measurements (multiple scans) Integration times for each scan Calibration Lamp Scans Aux/Cal Scans Aux/Empty Scans

36 Summary: Uncertainty Sources Where do I get my Estimates? Existing Literature Measure it Spec Sheets Derive it Things to Look out for Does the literature data apply to my system? What am I capturing in the measurements? Does my measurement isolate the sensitivity of terms? Is Spec sheet information sufficient? Did I validate my derivations? What are my assumptions

37 Sphere Calibration Coefficient, Counts/nm/ms/W Expanded Uncertainty.% Summary: Sample Output Uncertainty Budget at 550nm Component Description Symbol Value Unit Standard Uncertainty, u(x i ) Type DOF, ni Sensitivity, c i Contribution, u i (y) 1 Spectral Flux Reference Standard fr,l(l) W/nm 3.631E-04 B 4.055E E+00 2 Reference Standard Scan y R (l)-y Rd (l) counts/nm/ms 4.845E-04 A E E-02 3 Empty Sphere Scan with Aux y E (l)-y Ed (l) counts/nm/ms 5.419E-03 A 2.701E E-02 4 Aux scan with Reference Standard y xrd (l)-y xrd (l) counts/nm/ms 2.325E-03 A 9.817E E-02 5 Reference Lamp Current J A 2.701E-03 B E E-02 6 Spectral Shift exponent n JR E-03 B E E-04 7 Lamp aging factor A R E-05 B 7.827E E-03 8 Additional Uncertainty Contributions C f counts/nm/ms/w 1.803E-03 B 7.827E E Sphere Calibration Coefficient, C f (l) Expanded Uncertainty, % 12.0% Wavelength: 550 C f (l) Expanded Uncertainty 2.90 Uncertainty, % 3.7% % 8.0% 6.0% 4.0% 2.0% % Wavelength (nm)